Alerting & notifications
Set up alerting
Add a channel first, then a rule. Then use severities, escalation and maintenance windows to keep the volume sane.
Add a channel before you add a rule. A rule with no channel still records alerts in the app and still shows on the Alerts page — it just never reaches you, which is indistinguishable from "nothing has gone wrong".
1. Add a notification channel
- 1Open Alerts and go to the Channels tab.
- 2Choose a type: Email, Slack, Discord or a generic Webhook.
- 3For Slack or Discord, paste an incoming webhook URL from that app. For email, enter the address.
- 4Save, then send a test so you know it arrives before you rely on it.
A generic webhook posts JSON to anything that accepts it, and you can supply your own payload template if the receiving system expects a particular shape.
2. Add a rule
Rules live on the Rules tab. There are two kinds:
| Rule type | Fires when | Checked |
|---|---|---|
| Device offline | A device has been unreachable for the number of minutes you set (5 by default). | Every minute |
| Metric threshold | A metric crosses the value you set — and, if you ask, stays across it for a sustained period. | Every 5 minutes |
On a threshold rule, the sustained-time setting is the single most useful knob for cutting noise. A gap in the data resets the timer rather than counting as a breach, so a missed poll cannot page you on its own.
3. Keep the volume down
- Flap suppression — a device bouncing up and down is collapsed into one "flapping" alert instead of twenty up/down pairs, and stays collapsed until it holds steady.
- Severity — every alert carries a level, the Alerts page filters by it, and you can change the level on an individual alert.
- Escalation policies — on the Escalations tab: notify one channel, then another if nobody acknowledges. Acknowledging an alert stops further escalation.
- Maintenance windows — on the Maintenance tab. Planned work does not need to page anyone.
A maintenance window suppresses notifications and is subtracted from availability on your SLA reports. It is not subtracted from the availability figure on the dashboard, which counts wall-clock time. The two numbers can therefore differ for the same period, and the report is the one to quote to a customer — see how uptime is calculated.
Related
Did this miss something? Tell us — support is a small team and the fastest fix is usually a better article.